AIVResearch / MSANet

Official Pytorch implementation of Multi-Similarity and Attention Guidence for Boosting Few-Shot Segmentation.
https://arxiv.org/abs/2206.09667v1
63 stars 12 forks source link

Trsaining base model #7

Closed aminpdik closed 1 year ago

aminpdik commented 1 year ago

Hello

Thanks for the great work and for sharing the code.

I have one question regarding the training base model. Have you trained Resnet 101 base model considering the unseen classes as background same as BAM?

Thanks

Ehteshamciitwah commented 1 year ago

We use the pre-trained model provided by BAM. As the main focus was on the Meta Learner part. Yes, you are right the base learner is trained only on bass class assuming novel as background.

aminpdik commented 1 year ago

Thanks for your response.

BAM just provided weights for VGG and Resnet50. I think you trained in the same way as BAM for Resnet 101.

aminpdik commented 1 year ago

Hi @Ehteshamciitwah

Can I get the answer to my last question? BAM only provided weight for RESnet50 but you also provided RESnet101. Did you train the same as BAM for resnet101?

Best

Ehteshamciitwah commented 1 year ago

Sorry for the delay. yes, we trained base learner for ResNet 101 following BAM. The weights of ResNet101 are also provided.

aminpdik commented 1 year ago

Thank you for your answer